Sequentially partitioned population attributable fraction as an estimate of impact of tick presence on infection

Abstract

The impact of Rhipicephalus appendiculatus tick presence (exposure) on Theileria parva infection seroprevalence (outcome) on a farm was assessed in a group of cattle using population attributable fractions (PAF). The analysis was based on a representative sample of 80 traditional smallholder mixed farms from Mbeere District, Kenya. The PAFs were estimated using sequentially partitioned PAF approach that estimated a PAF associated with the exposure after adjusting for any effect that the confounder (agro-ecological zone [AEZ]) may have had by influencing the prevalence of the exposure. The resultant PAF was compared with Bruzzi approach PAF that estimated the proportion of T. parva infection cases directly attributable to the exposure after controlling for confounding by AEZ. The estimated PAF on the Bruzzi approach was 26.4% [95% CI: 9.6%, 43.2%]) whereas the partitioned PAF was 15.5% [95% CI: 1.5%, 29.6%]) implying that about 11% of the estimated impacts was driven by AEZ effects. Both approaches were consistent in estimating a relatively low impact of farm vector tick presence with a relatively high level of uncertainty. Overall, the results suggested that under endemic instability in Mbeere District, (1) presence of R. appendiculatus was not a good indicator of T. parva infection occurrence on a farm, and (2) ecological variation could play a role in determining infection impacts. This study provides a preliminary basis for evaluating the potential value of estimating PAFs for variables amenable to control in tick-borne diseases epidemiological studies.